采用复合特征提取和SVM的三电平STATCOM故障诊断方法  

Fault diagnosis for a three-level STATCOM based on the hybrid feature extraction and SVM

作　　者：李浩洋[1] 郭源博[1] 张晓华[1] LI Hao-yang;GUO Yuan-bo;ZHANG Xiao-hua(School of Electrical Engineering,Dalian University of Technology,Dalian 116023,China)

机构地区：[1]大连理工大学电气工程学院,辽宁大连116023

出　　处：《电机与控制学报》2019年第2期53-61,共9页Electric Machines and Control

基　　金：国家自然科学基金(51377013;51407023);中国博士后科学基金(2013M540224)

摘　　要：针对中点钳位式三电平静止同步补偿器(STATCOM)功率开关管的开路故障诊断问题,提出了一种复合故障特征提取技术。通过分析三电平STATACOM功率开关管开路故障后输出电流的变化规律,在小波包能量谱特征提取法的基础上,加入归一化平均电流特征提取法,计算一个周期内三相电流的归一化平均值,从而得到频域和时域两方面故障信号指标,构造故障特征向量;采用支持向量机对故障进行分类,通过改变负载来获取训练样本进而采用网格搜索法训练支持向量机。分别采用Matlab离线仿真和d SPACE实时仿真对所提算法进行验证,实验结果表明,与单一小波包能量谱特征提取技术相比,采用复合特征提取技术,故障诊断的准确度分别提高了38. 88%和49. 43%,诊断时间缩短了7. 38%和6. 61%。Focusing on the issue of open-circuit fault diagnosis of power switches in the neutral point clamped(NPC)three-level STATCOM,this paper proposes a hybrid fault feature extraction technique.By analyzing the change law of output current after power switch open-circuit fault of the three-level STATACOM,the normalized average current feature extraction method was applied on the basis of wavelet packet energy spectrum feature extraction method.The normalized average values of the three-phase currents were calculated in one cycle.Fault signal indicators were extracted from both frequency domain and time domain,and then the fault feature vectors were constructed.Support vector machine(SVM)was adopted to classify the faults.The training samples were obtained by changing the load and then the SVM was trained by using grid search method.The proposed technique was verified by using Matlab offline simulations and dSPACE real-time simulations,respectively.The results show that comparing with the single wavelet packet energy spectrum feature extraction technique,the hybrid feature extraction technique increases the accuracy of fault diagnosis by 38.88% and 49.43%,respectively,and reduces the diagnostic time by 7.38%and 6.61%,respectively.

关 键 词：三电平STATCOM 故障诊断 复合特征提取 支持向量机 dSPACE实时仿真 

分 类 号：TM71[电气工程—电力系统及自动化]